concurring:
I agree that MacPherson is not entitled to withdraw his guilty plea because not only is there no plain error, as the Court holds, there is no error at all. I add these additional views to explain why this is so, a discussion that endeavors to resolve the tension between United States v. Palladino, 347 F.3d 29 (2d Cir.2003), and United States v. Habbas, 527 F.3d 266 (2d Cir. 2008), this Court’s two most recent decisions on plea withdrawal claims based on a prosecutor’s advocacy at sentencing of a Sentencing Guidelines range higher than the range estimated in a plea bargain. These decisions differ not only in their outcomes (plea agreement violated in Palladino; plea agreement not violated in Habbas), but critically in their rationale as to what circumstances establish a breach of a plea agreement based on a so-called Pimentel estimate. See United States v. Pimentel, 932 F.2d 1029, 1034 (2d Cir.1991).
Notwithstanding these differences, the decisions are similar in several respects. In both Palladino and Habbas, the plea agreement included an estimate by the prosecutor as to the applicable sentencing range under the Sentencing Guidelines. Both agreements also included a statement that the estimate was based on information known to the Government at the time of the plea agreement. And in both cases, the prosecutor contended at sentencing that the applicable Guidelines sentencing range was higher than originally estimated. Despite these similarities, we ruled that denying withdrawal of the plea was error in Palladino and was not error in Habbas.
Palladino found the agreement breached because “[i]t was ... logical for defendant to believe that the estimate, and the Government’s stance at the sentencing hearing, would not be altered in the absence of new information.... ” 347 F.3d at 34. The basis for that belief was said to be the phrase “based on information known to the Office at this time,” a phrase that preceded the Government’s Guidelines range estimate. Id. at 31.
There can be no doubt that Palladino considered this phrase critical. First, it was italicized when quoted. See id. Second, the Court noted as “a curious fact” that this phrase was not brought to the attention of the sentencing judge. See id. at 33. Third, and most important, the Court explicitly relied on the phrase in stating its rationale:
*221In the circumstances presented in this case, we believe that defendant had a reasonable expectation that the Government would not press the Court for an enhanced offense level in the absence of new information. The language of the agreement specifically stated that the Government’s estimate was “based on information known to the [Government] at [the time of the plea].” ... It was thus logical for defendant to believe that the estimate, and the Government’s stance at the sentencing hearing, would not be altered in the absence of new information....
Id. at 34.
However, Habbas views the phrase “known to the Government” as of little, if any, consequence:
The problem in Palladino was not that those words appeared in the Pimentel estimate, but rather the combination of passages of the plea agreement conferring assurance that the government would not advocate for a sentence higher than the estimate, with the aura of unfair dealing that underlay the government’s change of position. Based on our reading of Palladino, we very much doubt that the result of that case would have changed if, all else remaining the same, those words had not appeared in the Pimentel estimate.
Habbas, 527 F.3d at 272 n. 1. I agree with Habbas on this point. Whether or not the Government says that its estimate is based on information “known to the Government at this time,” the estimate could only be based on such information. It could not be based on what the Government does not know at that time!
The decisions also differ in the significance they attach to the phrases in both plea agreements stating that (a) the estimate “is not binding on the [United States Attorney’s] Office,” and (b) if the Guidelines level “advocated by the [United States Attorney’s] Office” is “different from the estimate,” the defendant could not withdraw the plea. Habbas says that these statements “warned in several different ways that the government was likely to advocate for a higher sentence.” Id. at 270-71. Palladino accorded no significance to these statements and ruled that the plea agreement was violated when the Government advocated an enhancement based on information that had been known at the time of the plea.
In light of Palladino, a defendant in Habbas, named Rahman, contended that whenever the Government makes a Pimentel estimate, the plea agreement prohibits the Government from deviating from the estimate in the absence of new evidence. However, Habbas explicitly rejected that interpretation:
[W]e reject Rahman’s argument that the government violated his rights under the plea agreement by advocating a higher Guidelines level than it had estimated in the plea agreement.
Habbas, 527 F.3d at 270. In addition, Habbas explicitly rejected the claim that Palladino “established a broad rule, categorically prohibiting the government from deviating from a Pimentel estimate, absent newly discovered facts.” See id. at 272.
The plea agreement in the pending case does not have the “known to the Office” language, which Palladino deemed critical, but which Habbas deemed inconsequential. On the other hand, the plea agreement does have the “not binding” and “advocate!]” language, which Habbas deemed critical, but which Palladino deemed inconsequential.
Palladino and Habbas are problematic, not only for their inconsistency but also for the uncertainty they risk for the use of Pimentel estimates. I agree with Habbas *222that there should not be a bright-line rule permitting a plea to be withdrawn whenever the Government advocates a sentence above the plea agreement estimate. If we start inquiring as to what the Government knew at the time of the plea agreement, we open up a difficult area. Would the Government “know” of more offense conduct at the time of the plea agreement only if the evidence was in the office, or if an agent had told a prosecutor about it, or if a confidential informant had told the agent about it? How reliable would the information have to be to be “known” to the Government for purposes of making a Pimentel estimate binding? In the pending case, the Government’s knowledge of the heroin at the time of the plea agreement is clear, but if we were to make withdrawal of a plea available whenever the Government can be said to have “knowledge” of more offense conduct than is reflected in a Pimentel estimate, we would open up a potentially broad inquiry of uncertain limits.
I think a far better approach is to uphold all plea agreements with Pimentel estimates, regardless of whether the Government at sentencing advocates a higher Guidelines range (even one based on previously known facts), as long as the agreement makes clear that the Government is not bound by the estimate and the district judge ascertains at the plea colloquy that (1) the defendant understands that the estimate is not binding and (2) if the estimate is wrong, the defendant will not be permitted to withdraw his plea. That would leave Pimentel letters where they were initially intended to be: only estimates and not a basis for limiting the Government’s sentencing advocacy nor a basis for withdrawing a plea.
In this case, the plea agreement explicitly stated:
The Guidelines estimate put forth in paragraph 2 is not binding on the Office, the Probation Department or the Court. If the Guidelines offense level advocated by the Office, or determined by the Probation Department or the Court, is different from the estimate, the defendant will not be entitled to withdraw his plea.
Plea Agreement, ¶ 3. This language makes clear that the Government may advocate at sentencing a Guidelines offense level than the estimate. The agreement in Habbas contained identical language, see Habbas, 527 F.3d at 270, and the Habbas opinion relied on it, see id.
More important, in our ease, the plea colloquy included the following:
THE COURT: The sentencing when we before, and I hear Government, we will guidelines will be depart from them or bottom line is that until the date of get a presentence report, as I said from you, your lawyer and from the not know with any certainty what the or whether there will be grounds to whether the Court will impose a non-guideline sentence, do you understand that?
THE DEFENDANT: Yes, Your Honor.
THE COURT: Do you understand that these are all estimates that are not binding on the government, Probation or the Court?
THE DEFENDANT: Yes, Your Honor.
THE COURT: Do you also understand that if this estimate is wrong, that you will not be permitted to withdraw your plea of guilty?
THE DEFENDANT: Yes, Your Honor.
Thus, the Court’s colloquy informed MacPherson that (a) until the Court heard from the Government, there would not be certainty as to what the applicable guide*223line would be, (b) the estimate in the plea agreement was not binding on the Government, and (c) if the estimate was incorrect, the plea could not be withdrawn.
A plea colloquy can be examined to determine a defendant’s understanding of a plea agreement. See United States v. Woods, 581 F.3d 531, 534 (7th Cir.2009); United States v. Azure, 571 F.3d 769, 773-74 (8th Cir.2009); United States v. Woolley, 123 F.3d 627, 632 (7th Cir.1997). The colloquy in this case removes any basis for permitting the defendant to withdraw his plea simply because, at the time of the plea agreement, the Government knew of facts that would have justified a higher estimate.
If we permit withdrawal of a plea in cases such as this, we risk two adverse consequences. First, the Government will likely stop making Pimentel estimates. These estimates are not required, and will not be continued if they serve as a frequent basis for post-sentencing plea withdrawals. Second, the defendant gets two bites at the apple: he first argues at sentencing for a lenient sentence, and if he does not get one, he then appeals on the ground that he should be allowed to withdraw his plea. A properly worded plea agreement and a clear plea colloquy concerning that agreement, both of which were present in the pending case, should avoid both consequences.